Elamur Research⌕
☰
home / papers / 2026 / Graph-Augmented LLMs for Swiss MP Ideology Prediction

Graph-Augmented LLMs for Swiss MP Ideology Prediction

Paper · 2026 · Yifei Yuan, Luis Salamanca, Sophia Schlosser et al.

Summary

Political scientists need to place Members of Parliament on an ideological scale. Recent LLM-based estimators mostly treat each MP as an isolated source of text and overlook the relational structure of parliament, such as committees, party groups and co-sponsored proposals. The authors present PG-RAG, a retrieval-augmented pipeline that queries a Swiss political knowledge graph and adds the retrieved subgraph to the LLM prompt alongside basic MP metadata. The graph is built from the Bulletin of the Swiss Parliament. They define three subgraph types: speech-centric, MP-centric and pursuit-centric. Each subgraph is passed to the model in one of two forms, either as an LLM-written summary of serialized triplets or as a raw list of nodes and links. Compared with zero-shot and few-shot LLM baselines of various sizes, graph context consistently improves ideology predictions, and the smaller model benefits most. Left-leaning MPs, especially Social Democrats, remain hard to place because models tend to predict moderate scores for them. The study shows that domain-specific relational knowledge can strengthen computational models of legislative behavior.

Key points

  • The knowledge graph covers legislative periods 48–51 (2007–2023). It is published by Brandenberger et al. (2024) and its construction is described in Salamanca et al. (2024). Its core entities include Person, Pursuit, Speech, Party, Committee and Canton.
  • The six variants combine three subgraph types with two encodings: SP-S/R, MP-S/R and PR-S/R. SP is purely textual, MP spans 1-hop institutional relations, and PR goes up to 2 hops through sponsored and co-sponsored Pursuits.
  • Reported gains across RAG settings are 16.4–22.4% for the smaller Qwen-8B and 8.7–16.5% for GPT-5.
  • The pursuit-based setting worked best for Qwen-8B but gave GPT-5 its smallest gain (8.7%). This suggests larger models depend less on external structure.
  • In the case study, a Social Democrat MP has a ground-truth score of 9.0. PG-RAG (MP-S) predicted 7.1, closer than GPT-5's 6.5.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓ETH ZurichZurich · Switzerland84 works · 9,181 citations🎓University of ZurichZurich · Switzerland80 works · 12,260 citations🎓Zurich University of Teacher EducationZurich · Switzerland3 works · 5 citations🎓Swiss Data Science CenterZurich · Switzerland2 works · 1 citation

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksSocial Sciences · 1,954 worksGeneral Social Sciences · 254 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Yifei Yuan, Luis Salamanca, Sophia Schlosser and Laurence Brandenberger. 2026. Graph-Augmented LLMs for Swiss MP Ideology Prediction. Proceedings of the 11th Edition of the Swiss Text Analytics Conference.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{yuan2026graph,
  title = {{Graph-Augmented LLMs for Swiss MP Ideology Prediction}},
  author = {Yifei Yuan, Luis Salamanca, Sophia Schlosser and Laurence Brandenberger},
  booktitle = {Proceedings of the 11th Edition of the Swiss Text Analytics Conference},
  year = {2026},
  url = {https://aclanthology.org/2026.swisstext-1.12/}
}

References

53 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Project Apertus, Alejandro Hernández-Cano, Alexander Hägele, Allen Hao Huang, Angelika Romanou, Antoni-Joan Solergibert, Barna Pasztor, Bettina Messmer, Dhia Garbaya, Eduard Frank urech, Ido Hakimi, Juan García Giraldo, Mete Ismayilzada, Negar Foroutan, Skander Moalla, Tiancheng Chen, Vinko Sabolec,… — Apertus: Democratizing Open and Compliant LLMs for Global Language Environments
  2. Michael A Bailey and Erik Voeten. 2018. A two-dimensional analysis of seventy years of united nations voting. Public Choice, 176(1):33–55. find
  3. Ryan Bakker, Catherine De Vries, Erica Edwards, Liesbet Hooghe, Seth Jolly, Gary Marks, Jonathan Polk, Jan Rovny, Marco Steenbergen, and Milada Anna Vachudova. 2015. Measuring party positions in europe: The chapel hill expert survey trend file, 1999– 2010. Party Politics, 21(1):143–152. find
  4. Michael Barber. 2022. Comparing campaign finance and vote-based measures of ideology. The Journal of Politics, 84(1):613–619. find
  5. Pietro Bernardelle, Leon Fröhling, Stefano Civelli, Riccardo Lunardi, Kevin Roitero, and Gianluca Demartini. 2024. Mapping and influencing the political ideology of large language models using synthetic personas. Companion Proceedings of the ACM on Web Conference 2025. find
  6. Laurence Brandenberger, Julian Minder, Luis Salamanca, Sophia Schlosser, Lilian Gasser, Vincent Jung, Kourosh Shariat, Marta Balode, Anna Schmidt-Rohr, Leon Babi, Fernando Perez-Cruz, and Frank Schweitzer. 2024. DemocraSci - a parliamentary Knowledge Graph (4 legislative periods). find
  7. Laurence Brandenberger, Sophia Schlosser, Luis Salamanca, Lilian Gasser, Marta Balode, Julian Minder, Vincent Jung, Yaren Durgun, Leon Babic, Fernando Perez-Cruz, and Frank Schweitzer. 2026. DemocraSci: A Knowledge Graph on the Swiss parliament. Submitted Manuscript. find
  8. Michael Burnham. 2024. Semantic scaling: Bayesian ideal point estimates with large language models. arXiv preprint arXiv:2405.02472. — [arXiv:2405.02472](https://arxiv.org/abs/2405.02472) find
  9. Li Cai, Kilchan Choi, Mark Hansen, and Lauren Harrell. 2016. Item response theory. Annual Review of Statistics and Its Application, 3(1):297–321. find
  10. Clifford Carrubba, Matthew Gabel, and Simon Hug. 2008. Legislative voting behavior, seen and unseen: A theory of roll-call vote selection. Legislative Studies Quarterly, 33(4):543–572. find
  11. Clifford J Carrubba, Matthew Gabel, Lacey Murrah, Ryan Clough, Elizabeth Montgomery, and Rebecca Schambach. 2006. Off the record: Unrecorded legislative votes, selection bias and roll-call vote analysis. British Journal of Political Science, 36(4):691– 704. find
  12. R Philip Chalmers. 2012. mirt: A multidimensional item response theory package for the r environment. Journal of statistical Software, 48:1–29. find
  13. Yuxin Chen, Peng Tang, Weidong Qiu, and Shujun Li. 2025. Using llms for automated privacy policy analysis: Prompt engineering, fine-tuning and explainability. ArXiv, abs/2503.16516. find
  14. Yun-Shiuan Chuang, Ruixuan Tu, Chengtao Dai, Smit Vasani, Binwei Yao, Michael Henry Tessler, Sijia Yang, Dhavan Shah, Robert Hawkins, Junjie Hu, and 1 others. 2025. Debate: A large-scale benchmark for role-playing llm agents in multi-agent, long-form debates. arXiv preprint arXiv:2510.25110. — [arXiv:2510.25110](https://arxiv.org/abs/2510.25110) find
  15. Joshua Clinton, Simon Jackman, and Douglas Rivers. 2004. The statistical analysis of roll call data. American Political Science Review, pages 355–370. find
  16. Gary W Cox and Keith T Poole. 2002. On measuring partisanship in roll-call voting: The us house of representatives, 1877-1999. American Journal of Political Science, pages 477–489. find
  17. Kristina Dzeparoska, Jieyu Lin, Ali Tizghadam, and Alberto Leon-Garcia. 2023. Llm-based policy generation for intent-based management of applications. In 2023 19th International Conference on Network and Service Management (CNSM), pages 1–7. IEEE. find
  18. Muhammad Haroon, Magdalena Wojcieszak, and Anshuman Chhabra. 2025. "whose side are you on?" estimating ideology of political and news content using large language models and few-shot demonstration selection. In IJCNLP-AACL. find
  19. Wenyu Huang, Guanchen Zhou, Mirella Lapata, Pavlos Vougiouklis, Sébastien Montella, and Jeff Z. Pan. 2024. Prompting large language models with knowledge graphs for question answering involving long-tail facts. Knowl. Based Syst., 324:113648. find
  20. Simon Hug. 2010. Selection effects in roll call votes. British Journal of Political Science, 40(1):225–235. find
  21. Simon Hug and Tobias Schulz. 2007. LeftRight positions of political parties in Switzerland. Party Politics, 13(3):305–330. find
  22. Seth Jolly, Ryan Bakker, Liesbet Hooghe, Gary Marks, Jonathan Polk, Jan Rovny, Marco Steenbergen, and Milada Anna Vachudova. 2022. Chapel hill expert survey trend file, 1999–2019. Electoral studies, 75:102420. find
  23. Junsol Kim, James Evans, and Aaron Schein. 2025. Linear representations of political perspective emerge in large language models. ArXiv, abs/2503.02080. — [arXiv:2503.02080](https://arxiv.org/abs/2503.02080) find
  24. Peter E. Kraft, Hirsh Jain, and Alexander M. Rush. 2016. An embedding model for predicting roll-call votes. In EMNLP 2016 - Conference on Empirical Methods in Natural Language Processing, Proceedings. find
  25. Andreas Ladner. 2012. Switzerland’s green liberal party: a new party model for the environment? Environmental politics, 21(3):510–515. find
  26. Andreas Ladner. 2019. Switzerland: The" green" and" alternative parties". In New Politics In Western Europe, pages 155–165. Routledge. find
  27. Benjamin E Lauderdale and Alexander Herzog. 2016. Measuring political positions from legislative speech. Political Analysis, 24(3):374–394. find
  28. Michael Laver, Kenneth Benoit, and John Garry. 2003. Extracting policy positions from political texts using words as data. American Political Science Review, 97(02):311–331. find
  29. Hao Li, Viktor Schlegel, Yizheng Sun, Riza Theresa Batista-Navarro, and Goran Nenadic. 2025. Large language models in argument mining: A survey. ArXiv, abs/2506.16383. — [arXiv:2506.16383](https://arxiv.org/abs/2506.16383) find
  30. Yingjie Li, Tiberiu Sosea, Aditya Sawant, Ajith Jayaraman Nair, Diana Inkpen, and Cornelia Caragea. 2021. P-stance: A large dataset for stance detection in political domain. In Findings.
  31. Ruibo Liu, Chenyan Jia, Jason Wei, Guangxuan Xu, Lili Wang, and Soroush Vosoughi. 2021. Mitigating political bias in language models through reinforced calibration. ArXiv, abs/2104.14795. find
  32. Yujian Liu, Xinliang Frederick Zhang, David Wegsman, Nick Beauchamp, and Lu Wang. 2022. Politics: Pretraining with same-story article comparison for ideology prediction and stance detection. ArXiv, abs/2205.00619.
  33. Neo4j Core Team. 2024. Neo4j - the worlds leading graph database. find
  34. Lata Pangtey, Anukriti Bhatnagar, Shubhi Bansal, Shahid Shafi Dar, and Nagendra Kumar. 2025. Large language models meet stance detection: A survey of tasks, methods, applications, challenges and future directions. ArXiv, abs/2505.08464. find
  35. Pallavi Patil, Kriti Myer, Ronak Zala, Arpit Singh, Sheshera Mysore, Andrew McCallum, Adrian Benton, and Amanda Stent. 2019. Roll call vote prediction with knowledge augmented models. In CoNLL 2019 - 23rd Conference on Computational Natural Language Learning, Proceedings of the Conference. find
  36. Keith T Poole and Howard Rosenthal. 1985. A spatial model for legislative roll call analysis. American Journal of Political Science, 29(2):357–384. find
  37. Sven-Oliver Proksch and Jonathan B Slapin. 2010. Position taking in European Parliament speeches. British Journal of Political Science, 40(03):587– 611. find
  38. Luca Rettenberger, Markus Reischl, and Mark Schutera. 2024. Assessing political bias in large language models. Journal of Computational Social Science, 8. find
  39. Ludovic Rheault and Christopher Cochrane. 2020. Word embeddings for the analysis of ideological placement in parliamentary corpora. Political Analysis, 28(1):112–133. find
  40. Giuseppe Russo, Christoph Gote, Laurence Brandenberger, Sophia Schlosser, and Frank Schweitzer. 2023. Helping a friend or supporting a cause? disentangling active and passive cosponsorship decisions in the u.s. congress. ACL2023, The 61st Annual Meeting of the Association for Computational Linguisti… find
  41. Toronto, Canada, July 2023. find
  42. Luis Salamanca, Laurence Brandenberger, Lilian Gasser, Sophia Schlosser, Marta Balode, Vincent Jung, Fernando Perez-Cruz, and Frank Schweitzer. 2024. Processing large-scale archival records: The case of the Swiss parliamentary records. Swiss Political Science Review, 30(2):140–153. find
  43. Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. 2023. Whose opinions do language models reflect? ArXiv, abs/2303.17548.
  44. Eitan Sapiro-Gheiler. 2019. Examining political trustworthiness through text-based measures of ideology. In AAAI Conference on Artificial Intelligence. find
  45. Aaditya K. Singh, Adam Fry, Adam Perelman, Adam Tart, Adithya Ganesh, Ahmed El-Kishky, Aidan McLaughlin, Aiden Low, AJ Ostrow, Akhila Ananthram, Akshay Nathan, Alan Luo, Alec Helyar, Aleksander Madry, Aleksandr A Efremov, Aleksandra Spyra, Alex Baker-Whitcomb, Alex Beutel, Alex Karpenko, and 464 oth… — OpenAI GPT-5 System Card
  46. Barea M. Sinno, Bernardo Oviedo, Katherine Atwell, Malihe Alikhani, and Junyi Jessy Li. 2022. Political ideology and polarization: A multi-dimensional approach. In North American Chapter of the Association for Computational Linguistics. find
  47. Jonathan B Slapin and Sven-Oliver Proksch. 2008. A scaling model for estimating time-series party positions from texts. American Journal of Political Science, 52(3):705–722. find
  48. James M Snyder Jr and Tim Groseclose. 2000. Estimating party influence in congressional roll-call voting. American Journal of Political Science, 44(2):193– 211. find
  49. Keyon Vafa, Suresh Naidu, and David M Blei. 2020. Text-based ideal points. arXiv preprint arXiv:2005.04232. find
  50. Stefan Sylvius Wagner, Maike Behrendt, Marc Ziegele, and Stefan Harmeling. 2024. The power of llm-generated synthetic data for stance detection in online political discussions. ArXiv, abs/2406.12480. — [arXiv:2406.12480](https://arxiv.org/abs/2406.12480) find
  51. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. Qwen3 technical report.
  52. Tianyi Zhang. 2025. Probing political ideology in large language models: How latent political representations generalize across tasks. In Conference on Empirical Methods in Natural Language Processing. find
  53. Jianan Zhao, Le Zhuo, Yikang Shen, Meng Qu, Kai Liu, Michael M. Bronstein, Zhaocheng Zhu, and Jian Tang. 2023. Graphtext: Graph reasoning in text space. ArXiv, abs/2310.01089. — [arXiv:2310.01089](https://arxiv.org/abs/2310.01089) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yuan-2026-graph-augmented-llms-swiss-mp/